This paper presents a method of formalized answer extraction based on pattern learning.Our method extracts answers directly by matching the question patterns and the answer patterns automatically

which is different from the other traditional feature-based ones.The method gains formalized patterns automatically by using a machine learning strategy.Therefore

huge work load and low coverage of manually methods are avoided.Creatively

we apply logistic regression to estimate the confidence of the learned patterns.The contrast experiments give our method encouraging results

and it achieves fair and objective test in TREC QA track

the result of which shows that the method is a supplementary to the traditional feature based answer-extracting approaches.